Automated server workload management using machine learning

Patent metadata

Field	Value
Publication number	US-2018139271-A1
Application number	US-201815870262-A
Country	US
Kind code	A1
Filing date	Jan 12, 2018
Priority date	Oct 29, 2015
Publication date	May 17, 2018
Grant date	—

Abstract

Systems and methods are disclosed for managing workload among server clusters is disclosed. According to certain embodiments, the system may include a memory storing instructions and a processor. The processor may be configured to execute the instructions to determine historical behaviors of the server clusters in processing a workload. The processor may also be configured to execute the instructions to construct cost models for the server clusters based at least in part on the historical behaviors. The cost model is configured to predict a processor utilization demand of a workload. The processor may further be configured to execute the instructions to receive a workload and determine efficiencies of processing the workload by the server clusters based at least in part on at least one of the cost models or an execution plan of the workload.

First claim

1 - 20 . (canceled)

21 . A system for managing a workload among server clusters, comprising:
a memory unit storing instructions; and one or more processors configured to execute the stored instructions to perform operations comprising:
receiving a workload to be processed by one or more server clusters;
storing metadata of more than one server cluster on one or more memory units;
determining, based on the stored metadata, candidate server clusters to process the workload;
sending a query to more than one candidate server cluster for evaluating a processing cost;
receiving a processing cost from more than one candidate server cluster based on the query;
distributing the workload to a selected server cluster based on one or more predetermined conditions and on the received processing costs; and
receiving a processing result from the selected server cluster.

22 . The system of claim 21 , the operations further comprising:
parsing the workload into one or more component tasks.

23 . The system of claim 22 :
wherein distributing the workload further comprises distributing the component tasks to a plurality of selected server dusters; and wherein the operations further comprise: receiving a processing result of more than one selected server duster; and
aggregating the received processing results into a final result.

24 . The system of claim 21 , wherein the one or more processors are configured as a distributed computer system including a plurality of computers that interoperate to perform the operations.

25 . The system of claim 21 , wherein each received processing cost is computed based on a respective cost model and a respective workload execution plan.

26 . The system of claim 25 , wherein each received processing cost is the lowest processing cost computed from among alternative plans for executing the workload on the respective server duster.

27 . The system of claim 25 , wherein at least one cost model is based on at least one of the historical behavior of the respective candidate server, the composition of hardware resources of the respective candidate server, and the capacity of the respective hardware resource.

28 . The system of claim 25 , where each cost model is constructed using machine learning algorithms.

29 . The system of claim 21 , wherein receiving the workload comprises receiving the workload from a user device.

30 . The system of claim 21 , the operations further comprising:
receiving updated metadata for one of the server dusters; and updating the stored metadata with the updated metadata

31 . The system of claim 21 , the operations further comprising:
determining an importance level of the workload; and wherein the distributing the workload to a selected server duster is further based on the importance level of the workload.

32 . The system of claim 31 , the operations further comprising:
directing the selected server duster to process the workload based on the importance level of the workload.

33 . The system, of claim 31 , wherein determining the importance level of the workload is based on a user input received by the one or more processors.

34 . The system of claim 21 , the operations further comprising:
distributing, by the one or more processors, the workload to a cloud service.

35 . The system of claim 21 , the operations further comprising:
intercepting, by the one or more processors, the received workload at an abstraction layer.

36 . The system of claim 21 , wherein each query includes an execution plan of the workload.

37 . The system of claim 21 , the operations further comprising at least one of:
changing, by the one or more processors, a priority of the workload; deferring, by the one or more processors, the workload for processing at a later time; or rejecting, by the one or more processors, the workload.

38 . The system of claim 21 , wherein the metadata comprises storage locations of data in each server duster and a current resource availability of each server duster.

39 . A method for managing a workload among server dusters, comprising:
receiving, by one or more processors, a workload to be processed by one or more server dusters; storing, by the one or more processors, metadata of more than one server cluster on one or more memory units; determining, by the one or more processors, based on the stored metadata, candidate server dusters to process the workload; sending, by the one or more processors, a query to more than one candidate server cluster for evaluating a processing cost; receiving, by the one or more processors, a processing cost from more than one candidate server duster based on the query; distributing, by the one or more processors, the workload to a selected server duster based on one or more predetermined conditions and on the received processing costs; and receiving, by the one or more processors, a processing result from the selected server duster.

40 . A non-transitory computer readable medium having stored instructions, which when executed, cause at least one processor to perform operations for managing a workload among server dusters comprising:
receiving a workload to be processed by one or more of server clusters; storing metadata of more than one server duster on one or more memory units; determining based on the stored metadata, candidate server dusters to process the workload; sending a query to more than one candidate server duster for evaluating a processing cost; receiving a processing cost from more than one candidate server duster based on the query; distributing the workload to a selected server cluster based on one or more predetermined conditions and on the received processing costs; and receiving a processing result from the selected server cluster

Assignees

• Capital One Services Llc

Inventors

• Kumar Subodh
• Bardwaj Santosh

Classifications

• G06F3/0608

  Saving storage space on storage systems · CPC title

• G06F9/505

  considering the load · CPC title

• G06F3/0619

  in relation to data integrity, e.g. data losses, bit errors · CPC title

• G06F9/5061

  Partitioning or combining of resources · CPC title

• G06F3/1211

  Improving printing performance · CPC title